0.1.0 beta

VoxtMini // For dictation that stays local

Speak clearly. Adapt to context.

Press Right ⌥ in any app and speak. Models on your Mac write it out at your cursor.

  • English, Chinese and Japanese
  • A remote service only where you pick one
  • Free and open source (Apache-2.0)

Who it's for //

For dictation that stays local

For
People who dictate in English, Chinese or Japanese and want their audio to stay on the Mac by default.
What it gives them
Voxt's features on local models, with fewer settings, and a remote service for any feature you choose.

At a glance //

What is VoxtMini?

VoxtMini is voice input for the Mac: press a shortcut in any app, speak, and the text appears where your cursor is. Local models are the default; a feature uses a remote service only when you choose one for it. English, Chinese and Japanese.

What it is
Voice input · Local AI · Open source
Price
Free and open source (Apache-2.0)
Runs on
macOS 15+ · Apple silicon
Languages
English and 简体中文
Latest version
0.1.0-beta.8 (beta)

By the numbers //

Measured numbers name the Mac they ran on · none from unreleased builds

3

languages: English, Chinese and Japanese

0

uploads unless you choose a remote service

Features //

Voice input on local models, remote by choice

Privacy // Where your data goes

Local unless you choose otherwise

By default audio and text never leave the Mac. A feature you set to a remote service sends its audio or text to that service only.

  1. 01Remote keys stay in the macOS Keychain.
  2. 02No account, no analytics, no server of VoxtMini's.
  3. 03Online only for model downloads, update checks and announcements.

Open source //

Source on GitHub

Open source under Apache-2.0.

View on GitHub ↗

Requirements

macOS
15 or later
Chip
Apple silicon
Memory
16 GB recommended; 8 GB with the fast models
Disk
About 6 GB for the recommended models
Languages
English and 简体中文
No account

Permissions

  • Microphone

    To record while a shortcut is active, and only then.

  • Accessibility

    To type results into other apps and read the selected text. Without it, results are copied instead.

FAQ

Questions

VoxtMini replaces Voxt on actnow.dev: Voxt rewritten in Go, with fewer settings, for English, Chinese and Japanese, and remote services as an option. Voxt itself continues in SGLang-Omni.

Latest release

0.1.0-beta.8

Nov 17, 2026 · beta

Full changelog

Fixed

  • Local models work on macOS 15 again: the speech runtime was built for macOS 26 only, so on macOS 15 every local speech model failed to start ("language version 4.0 which is not supported on this OS"). Releases are now checked so that nothing in them needs a newer macOS than 15.0. If this ever happens again, VoxtMini says that this build does not support the Mac's macOS version.

VoxtMini, open source

Open source under Apache-2.0. Download it above, or read the code on GitHub.